No idea why this is happening, the HTTP POST request to our internal FastAPI analytics service fails consistently during load testing, while it works perfectly in manual debug mode. The integration is deployed via Terraform genesyscloud_architect_flow resource, and the data action configuration appears valid in the UI.
The setup involves a Python FastAPI endpoint hosted on AWS ECS (Fargate) that processes call transcripts and returns sentiment scores. The Genesys Cloud flow triggers this endpoint using the “Make an HTTP Request” data action. Under normal single-threaded execution, the response time is ~200ms, and the status code is 200. However, when simulating concurrent calls via JMeter or during peak hours (AEST 09:00 - 11:00), the Genesys Cloud flow logs a 504 Gateway Timeout. The FastAPI logs show that the request never arrives at the ECS task, suggesting the drop-off happens at the Genesys Cloud outbound proxy layer or the connection pool is exhausted.
Environment details:
- Genesys Cloud Org ID: 12345678
- Architect Flow ID: 8a1b2c3d-4e5f-6789-0abc-def123456789
- Data Action Type: HTTP Request (POST)
- Target Endpoint: https://api.internal-analytics.com/v1/sentiment
- Timeout Setting: 30000 ms (default)
- Terraform Provider Version: 1.15.0
- FastAPI Version: 0.104.1
- Uvicorn Workers: 4
The Terraform configuration for the data action is minimal, relying on the UI for the endpoint URL to avoid hardcoding secrets. We have verified that the VPC endpoints are correctly configured and that the security group allows outbound traffic to the specific IP range of the Genesys Cloud data action servers. The issue is not related to payload size, as the JSON body is under 1KB.
Is there a known rate limit or connection pool constraint for outbound HTTP requests from Genesys Cloud data actions? We have checked the WFM documentation, but it does not specify limits for custom HTTP integrations. Any insight into debugging the 504 at the Genesys Cloud network edge would be appreciated.